Record Details

Replication data for: What to do about Atheoretic Lags

Harvard Dataverse (Africa Rice Center, Bioversity International, CCAFS, CIAT, IFPRI, IRRI and WorldFish)

View Archive Info
 
 
Field Value
 
Title Replication data for: What to do about Atheoretic Lags
 
Identifier https://doi.org/10.7910/DVN/29362
 
Creator Cranmer, Skyler
Rice, Douglas
Siverson, Randolph
 
Publisher Harvard Dataverse
 
Description We examine a problem that is confronted frequently by political science researchers seeking to model longitudinal data: what to do when one suspects a lag between the realization of a regressor and its effect on the outcome variable, but one has no theoretical reason to suspect a particular lag length. We examine the theoretical challenges posed by non-specific lags, review existing methods for non-specific lag analysis - most notably distributed lag specifications - and their shortcomings, and present an alternative approach for atheoretic lag analysis based on Bayesian model averaging (BMA). We demonstrate the use and utility of our approach with two examples: the litigant signal model in American politics and modernization theory in political economy. Our examples show the increasing difficulty of analyzing models with atheoretic lags as the set of possible specifications increases, and demonstrate the effectiveness of Bayesian model averaging for the modal type of specification in time-series cross-sectional applications.